Phase Grouping in PV-Rich LV Feeders: Smart Meter Data and Unconstrained k-Means

Abstract

The growing amount of residential solar photovoltaic (PV) systems has increased the need for adequate three-phase low-voltage (LV) feeder models so distribution companies can carry out detailed studies that identify potential impacts and solutions. One of the major challenges, however, is that the phase group to which residential customers are connected to is largely unknown. Using voltage time-series data extracted from single-phase smart meters (a technology being deployed around the world), this paper proposes an approach based on Principal Component Analysis (PCA) and an unconstrained k-means clustering technique to determine the phase group of customers in a given LV feeder. The approach, which does not require prior knowledge (i.e., phase group of a few customers), is tested on a realistic Australian LV feeder with 31 customers and different PV penetrations. Results demonstrate that the approach is accurate without and with PV and practical as only smart meter data is required.